Anthropic said it identified and removed three accounts aligned with Iranian state propaganda institutions
They also converted official government intelligence bulletins into tailored content in Farsi, Arabic, Urdu, Malay, Spanish and English, with plans for distribution in as many as 20 languages. Iranian state-aligned and Iran-linked groups used Anthropic’s Claude artificial intelligence platform to run propaganda campaigns, conduct domestic surveillance and develop targeting recommendations against US naval forces, according to a threat report released by the company.
The accounts used Claude to develop campaign plans, propaganda doctrine manuals, personas and target databases, according to Anthropic. The findings come amid heightened tensions between Iran and the United States, including attacks involving US and Iranian forces around the Strait of Hormuz. Commercial shipping through the waterway has remained limited as the conflict has continued, with Iran seeking greater control over traffic passing through the strategic strait.
The account used the AI model to compile targeting handbooks and identify and track naval positions using open-source information, according to the company. Anthropic said the activity involved developing recommendations that could potentially be used against US naval forces. Anthropic separately identified an Iran-linked account that used Claude to collect and analyse publicly available information about US naval forces in the region.